[PATCHES] Patches for the AD DC

[Andrew Bartlett]

On Mon, 2013-05-27 at 09:38 +1000, Andrew Bartlett wrote:
> On Fri, 2013-05-17 at 11:02 +1000, Andrew Bartlett wrote:
> > On Fri, 2013-05-17 at 09:23 +1000, Andrew Bartlett wrote:
> > > On Fri, 2013-05-17 at 00:40 +0200, Stefan (metze) Metzmacher wrote:
> > > > Hi Andrew,
> > > > 
> > > > > On 05/16/2013 04:45 AM, Andrew Bartlett wrote:
> > > > >> Attached is a series of patches I've been working on in various places,
> > > > >> all collected up.  I'm doing a private autobuild on these now, but I've
> > > > >> cut them down so that they pass a number of DRS tests that were a
> > > > >> problem before.
> > > > >>
> > > > >> In any case, some review of these would be nice.
> > > > >>
> > > > >> In particular, I now have code to handle and test (!) some more complex
> > > > >> rename situations that I think we were getting wrong before.
> > > > >>
> > > > >> Andrew Bartlett
> > > > > Patch dsdb-linked_attributes: Do not error when we cannot find the backlink
> > > > > 
> > > > > Can you add a debug message if we are not returning an error if the
> > > > > fowardlink has no backlink ?
> > > > 
> > > > I also thought that.
> > > > 
> > > > I'm also wondering in what situation this could happen?
> > > > Can you explain that?
> > > 
> > > It came up in my testing.  I'll re-add in the patches that triggered it
> > > and show you.  I couldn't understand it either, when I examined the
> > > databases at the end the objects that caused the crash were present, and
> > > we set all the magic 'really show me the objects' flags. 
> > 
> > It may well have only come up due to flaws in my original rename support
> > patch before I finished it.  I can't reproduce it any more.
> > 
> > I attach the two options:
> >  - give the error
> >  - ignore the error
> > 
> > I still think it is up to dbcheck to clean this up, not for runtime errors to be given
> > for the only loosely related task of doing the rename, but either way we
> > should ensure even in the 'impossible' situation, we don't segfault. 
> 
> Metze,
> 
> Where did we end up with this, and more importantly with the rename
> patch?

Can we at least merge the rename patch and tests, while we sort out the
linked attributes issue?

I would really like to make some progress here, and move on to other DRS
issues, like fixing dbcheck and linked attribute hiding, because I think
incorrect handling in dbcheck is giving us false positives that are
blocking our work here.

Andrew Bartlett

-- 
Andrew Bartlett                                http://samba.org/~abartlet/
Authentication Developer, Samba Team           http://samba.org